October 7. In the morning I pulled the register first.
The three lines I saw at five yesterday were as they were. The first line, 6 October 1999. The second, 6 October 99. The third, 20 October 1999. All three have the same material number and the same recipient.
I went down to the stacks. The one gap that was empty yesterday was as it was, and the place beside it was filled. The box the third line points to is still there.
At nine I ran the search. Two hundred and ninety-one.
Writing it into the list, I caught once. Yesterday's two are the same day written as two years, so duplicate entry is right. The third is a different day. It does not belong in the same list and there is nowhere else to put it. I put it there so as not to make a sixth list.
At ten I started on the index cards. Drawer seven.
Drawer six I finished yesterday. Six drawers in ten days is half. The remaining six will be done in the first days of November. That is twenty days ahead of the deadline.
What one does to a card is always the same. Look at the stamped two digits and write four beside them. I have done this to more than three thousand cards, so the hand goes first. In the morning I did a hundred and twenty.
At half past eleven I pulled the catalogue. I meant to see how a converted drawer comes up on it.
There were eleven lines with late October dates.
Having written future I struck it out. There is no such word in the form. I wrote registration date subsequent instead.
I opened the eleven one by one.
Nine were on the test machine. One is the test machine's catalogue program and eight are index updates held on that machine. Four are 20 October and four are 31 October.
The reason was in my log.
In the latter part of September further test guidance came down from head office. It was to see whether the corrected catalogue distributed on 6 September reads dates properly. Five dates were written in the guidance. 20 October, 31 October, 30 November, 31 December, 1 January 00. I was to move the test machine's clock to each of the five in turn, pull the catalogue, and confirm the sort.
I did it on 27 September. It is written on page one hundred and forty-one of the log. All five sorted correctly and I replied no anomaly.
While the clock stands moved, an index update on that machine has its update date entered as that day. The machine believes its own clock. That is why there are nine items for 20 October and 31 October.
The explanation fitted exactly. Not one of the nine dates departs from the dates in the guidance.
The remaining two were not on the test machine.
One was the operational machine's index. Its date is 31 October. The operational machine's clock has never once been moved. There was no guidance to move it, and moving it would enter that day's whole work under that date, so it cannot be moved.
The other was the release-register program.
The release register was converted to four digits in August. I converted it, and I did not touch the clock in doing so. There is no occasion to. The register takes the date from the machine's clock as it enters, so the clock must stand right for the entry to stand right.
At two I telephoned head office. It was not the person I spoke to yesterday.
I told him about the nine and he said they knew of it. It is a common remainder after a test, and when the conversion finishes at the end of November separate guidance for clearing them will come down.
I told him about the remaining two. He was silent a moment and then asked how many times I had run the test. Once, I said. He asked whether the test machine and the operational machine are joined by a line. No, I said. Material is carried between them by hand on a disk.
He said then I should check again when it was that the clock had been moved at our end.
At three the third-floor man came down and spoke about yesterday's box. Apart from its being entered twice, he asked, is there one entered as going out again on the twentieth. He had seen it looking at the catalogue upstairs.
There is, I said. He asked what he should hand out on the twentieth, then. If a requisition comes it is handed out, I said. He asked what happens if no requisition is made.
I do not know, I said. He laughed and went up.
After he went up I wrote the question down. For nine months what I have asked in this room has always been how a thing stands. What he asked is what one does. If the release entered in the register is not made the register is wrong, and if it is made two boxes must go out, and there is one box. Either way something impossible must be done thirteen days from now.
From half past three until five I went through the whole of the operational machine's records. Touching the clock leaves a record. There were six lines from January until today and all six are time corrections. A few seconds lost each day, put right month by month. The largest move was twelve seconds.
There was no line that moved a date.
Of the things nine months in this room have taught me, there is this. When an explanation comes, what lies outside it remains. In February the paper remained, in June the corrected text remained, in August three places remained. What remains is never outside the explanation but at the end of it. The explanation goes as far as nine and does not go from the tenth.
Today it went as far as nine. The nine agree with the dates in the guidance without one character out. In eight months this is the first explanation to fit so well.
That explanation cannot lay a finger on the remaining two. There is no place to lay it. The remainder of a test cannot come out of a machine that was not tested.
The better an explanation fits, the narrower the place it does not fit. Narrowed, it looks small. Two out of eleven is about two parts in twenty. In June, when sixteen out of twenty were wrong, I held that to be large. Today's two look small, and their looking small is welcome.
At five I understood what it was I was welcoming. I was glad that eighteen parts in twenty had been explained. What was explained did nothing today, and I was glad to be able to write that the nothing came to eighteen. Then what happened today is two. For eight months what happened in a day has always been one or two. Whether the nine are explained or not, what remains is two. An explanation does not reduce the number of things that have happened.
At ten past five I wrote the last line and closed the log.
At half past five I opened the access record.
I meant to see who had registered the eleven. Registering must be done by a person, and when a person does it an access record remains. There is one account in this room that can register.
Nine of them had no access record at all. The test machine leaves no access record. The tenth had none either.
The eleventh had one line.
The account was mine. The object of the query was '1997 Central Archive Release Register'. It is the one whose body would not open on 9 September.
The time was ten twelve a.m.
The date was 8 October.